Fang, Xuan and Varga, István and Tettamanti, Tamás (2026) Multi-Agent Reinforcement Learning-Based Variable Speed Limit Control for Mixed Urban Traffic Flow. IEEE ACCESS, 14. pp. 67991-68003. ISSN 2169-3536 10.1109/ACCESS.2026.3689744

Text Fang_67991_37215853_ny.pdf Download (3MB)

Abstract

With the acceleration of urbanization and the advent of autonomous driving technology, urban traffic systems are gradually evolving into a mixed traffic flow environment where Human Driven Vehicles (HAVs) and Connected and Automated Vehicles (CAVs) coexist. Conventional urban traffic control is primarily based on traffic lights with fixed or adaptive logic, as other means of actuation are not traditionally in use (except for tolling systems, which are rather indirect regulators). Traditional traffic signal control, however, is limited due to its rigidity; traffic lights are located at fixed spots with a strict operational mechanism (cycle time, phase order, offset), making them less efficient in managing a mixed traffic environment. To address this challenge, this paper follows a novel approach leveraging the Variable Speed Limit (VSL) control in the urban context. The control framework is based on Multi-Agent Reinforcement Learning (MARL), aiming to optimize road sustainability under mixed traffic flow conditions. This study employs a Centralized-Training Decentralized-Execution (CTDE) architecture, utilizing the Multi-Agent Proximal Policy Optimization (MAPPO) algorithm to address the curse of dimensionality and non-stationarity issues in large-scale urban road networks. This framework enables collaborative decision-making in both space and time through the use of shared parameters and global value function evaluation. The method is validated via realistic traffic simulation experiments in Simulation of Urban MObility (SUMO), demonstrating that the proposed control strategy can significantly reduce travel time, pollutant emissions, and improve traffic safety on the studied urban network under different traffic conditions.

Item Type:	Article
Subjects:	Q Science > QA Mathematics and Computer Science > QA75 Electronic computers. Computer science / számítástechnika, számítógéptudomány
Divisions:	Systems and Control Lab
SWORD Depositor:	MTMT Injector
Depositing User:	MTMT Injector
Date Deposited:	09 Sep 2026 11:14
Last Modified:	09 Sep 2026 11:14
URI:	https://eprints.sztaki.hu/id/eprint/11139

Update Item